DAG-GCN: Directed Acyclic Causal Graph Discovery from Real World Data using Graph Convolutional Networks

Se Joon Park; Jihie Kim

doi:10.1109/BigComp57234.2023.00065

DAG-GCN: Directed Acyclic Causal Graph Discovery from Real World Data using Graph Convolutional Networks

Se Joon Park, Jihie Kim

Department of Computer Science and Artificial Intelligence

Dongguk University

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

6 Scopus citations

Abstract

Causal discovery has been challenging since the search space of directed acyclic graphs super-exponentially grows with respect to the number of nodes. Previously constraint-based and score-based methods have been used. In recent studies, a continuous optimization method has reached a high score, but the problem is still harsh in real-world observational data. Motivated by the success of recent GNN models, we extended previous methods to be suitable to actual world data. Our model is based on the DAG-GNN model, uses GCN, and tries to learn an adjacency matrix set as a model parameter. To solve the vanishing adjacency matrix problem, we use the He-Initialization method with Leaky ReLU and the batch normalization technique. We demonstrate our model on real-world data sets. Compared to the state-of-the art results, our proposed method reaches acceptable results.

Original language	English
Title of host publication	Proceedings - 2023 IEEE International Conference on Big Data and Smart Computing, BigComp 2023
Editors	Hyeran Byun, Beng Chin Ooi, Katsumi Tanaka, Sang-Won Lee, Zhixu Li, Akiyo Nadamoto, Giltae Song, Young-guk Ha, Kazutoshi Sumiya, Wu Yuncheng, Hyuk-Yoon Kwon, Takehiro Yamamoto
Publisher	Institute of Electrical and Electronics Engineers Inc.
Pages	318-319
Number of pages	2
ISBN (Electronic)	9781665475785
DOIs	https://doi.org/10.1109/BigComp57234.2023.00065
State	Published - 2023
Event	2023 IEEE International Conference on Big Data and Smart Computing, BigComp 2023 - Jeju, Korea, Republic of Duration: 13 Feb 2023 → 16 Feb 2023

Publication series

Name	Proceedings - 2023 IEEE International Conference on Big Data and Smart Computing, BigComp 2023

Conference

Conference	2023 IEEE International Conference on Big Data and Smart Computing, BigComp 2023
Country/Territory	Korea, Republic of
City	Jeju
Period	13/02/23 → 16/02/23

Keywords

Causal Discovery
Causal Structure Learning
DAG
Directed Acyclic Graphs
Graph Neural Networks
Graph Representation Learning

Access to Document

10.1109/BigComp57234.2023.00065

Cite this

Park, S. J., & Kim, J. (2023). DAG-GCN: Directed Acyclic Causal Graph Discovery from Real World Data using Graph Convolutional Networks. In H. Byun, B. C. Ooi, K. Tanaka, S.-W. Lee, Z. Li, A. Nadamoto, G. Song, Y. Ha, K. Sumiya, W. Yuncheng, H.-Y. Kwon, & T. Yamamoto (Eds.), Proceedings - 2023 IEEE International Conference on Big Data and Smart Computing, BigComp 2023 (pp. 318-319). (Proceedings - 2023 IEEE International Conference on Big Data and Smart Computing, BigComp 2023). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/BigComp57234.2023.00065

Park, Se Joon ; Kim, Jihie. / DAG-GCN : Directed Acyclic Causal Graph Discovery from Real World Data using Graph Convolutional Networks. Proceedings - 2023 IEEE International Conference on Big Data and Smart Computing, BigComp 2023. editor / Hyeran Byun ; Beng Chin Ooi ; Katsumi Tanaka ; Sang-Won Lee ; Zhixu Li ; Akiyo Nadamoto ; Giltae Song ; Young-guk Ha ; Kazutoshi Sumiya ; Wu Yuncheng ; Hyuk-Yoon Kwon ; Takehiro Yamamoto. Institute of Electrical and Electronics Engineers Inc., 2023. pp. 318-319 (Proceedings - 2023 IEEE International Conference on Big Data and Smart Computing, BigComp 2023).

@inproceedings{817af7a5dc034e86a2766fcae18424ed,

title = "DAG-GCN: Directed Acyclic Causal Graph Discovery from Real World Data using Graph Convolutional Networks",

abstract = "Causal discovery has been challenging since the search space of directed acyclic graphs super-exponentially grows with respect to the number of nodes. Previously constraint-based and score-based methods have been used. In recent studies, a continuous optimization method has reached a high score, but the problem is still harsh in real-world observational data. Motivated by the success of recent GNN models, we extended previous methods to be suitable to actual world data. Our model is based on the DAG-GNN model, uses GCN, and tries to learn an adjacency matrix set as a model parameter. To solve the vanishing adjacency matrix problem, we use the He-Initialization method with Leaky ReLU and the batch normalization technique. We demonstrate our model on real-world data sets. Compared to the state-of-the art results, our proposed method reaches acceptable results.",

keywords = "Causal Discovery, Causal Structure Learning, DAG, Directed Acyclic Graphs, Graph Neural Networks, Graph Representation Learning",

author = "Park, {Se Joon} and Jihie Kim",

note = "Publisher Copyright: {\textcopyright} 2023 IEEE.; 2023 IEEE International Conference on Big Data and Smart Computing, BigComp 2023 ; Conference date: 13-02-2023 Through 16-02-2023",

year = "2023",

doi = "10.1109/BigComp57234.2023.00065",

language = "English",

series = "Proceedings - 2023 IEEE International Conference on Big Data and Smart Computing, BigComp 2023",

publisher = "Institute of Electrical and Electronics Engineers Inc.",

pages = "318--319",

editor = "Hyeran Byun and Ooi, {Beng Chin} and Katsumi Tanaka and Sang-Won Lee and Zhixu Li and Akiyo Nadamoto and Giltae Song and Young-guk Ha and Kazutoshi Sumiya and Wu Yuncheng and Hyuk-Yoon Kwon and Takehiro Yamamoto",

booktitle = "Proceedings - 2023 IEEE International Conference on Big Data and Smart Computing, BigComp 2023",

address = "United States",

}

Park, SJ & Kim, J 2023, DAG-GCN: Directed Acyclic Causal Graph Discovery from Real World Data using Graph Convolutional Networks. in H Byun, BC Ooi, K Tanaka, S-W Lee, Z Li, A Nadamoto, G Song, Y Ha, K Sumiya, W Yuncheng, H-Y Kwon & T Yamamoto (eds), Proceedings - 2023 IEEE International Conference on Big Data and Smart Computing, BigComp 2023. Proceedings - 2023 IEEE International Conference on Big Data and Smart Computing, BigComp 2023, Institute of Electrical and Electronics Engineers Inc., pp. 318-319, 2023 IEEE International Conference on Big Data and Smart Computing, BigComp 2023, Jeju, Korea, Republic of, 13/02/23. https://doi.org/10.1109/BigComp57234.2023.00065

DAG-GCN: Directed Acyclic Causal Graph Discovery from Real World Data using Graph Convolutional Networks. / Park, Se Joon; Kim, Jihie.
Proceedings - 2023 IEEE International Conference on Big Data and Smart Computing, BigComp 2023. ed. / Hyeran Byun; Beng Chin Ooi; Katsumi Tanaka; Sang-Won Lee; Zhixu Li; Akiyo Nadamoto; Giltae Song; Young-guk Ha; Kazutoshi Sumiya; Wu Yuncheng; Hyuk-Yoon Kwon; Takehiro Yamamoto. Institute of Electrical and Electronics Engineers Inc., 2023. p. 318-319 (Proceedings - 2023 IEEE International Conference on Big Data and Smart Computing, BigComp 2023).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - DAG-GCN

T2 - 2023 IEEE International Conference on Big Data and Smart Computing, BigComp 2023

AU - Park, Se Joon

AU - Kim, Jihie

PY - 2023

Y1 - 2023

N2 - Causal discovery has been challenging since the search space of directed acyclic graphs super-exponentially grows with respect to the number of nodes. Previously constraint-based and score-based methods have been used. In recent studies, a continuous optimization method has reached a high score, but the problem is still harsh in real-world observational data. Motivated by the success of recent GNN models, we extended previous methods to be suitable to actual world data. Our model is based on the DAG-GNN model, uses GCN, and tries to learn an adjacency matrix set as a model parameter. To solve the vanishing adjacency matrix problem, we use the He-Initialization method with Leaky ReLU and the batch normalization technique. We demonstrate our model on real-world data sets. Compared to the state-of-the art results, our proposed method reaches acceptable results.

AB - Causal discovery has been challenging since the search space of directed acyclic graphs super-exponentially grows with respect to the number of nodes. Previously constraint-based and score-based methods have been used. In recent studies, a continuous optimization method has reached a high score, but the problem is still harsh in real-world observational data. Motivated by the success of recent GNN models, we extended previous methods to be suitable to actual world data. Our model is based on the DAG-GNN model, uses GCN, and tries to learn an adjacency matrix set as a model parameter. To solve the vanishing adjacency matrix problem, we use the He-Initialization method with Leaky ReLU and the batch normalization technique. We demonstrate our model on real-world data sets. Compared to the state-of-the art results, our proposed method reaches acceptable results.

KW - Causal Discovery

KW - Causal Structure Learning

KW - DAG

KW - Directed Acyclic Graphs

KW - Graph Neural Networks

KW - Graph Representation Learning

UR - http://www.scopus.com/inward/record.url?scp=85151523717&partnerID=8YFLogxK

U2 - 10.1109/BigComp57234.2023.00065

DO - 10.1109/BigComp57234.2023.00065

M3 - Conference contribution

AN - SCOPUS:85151523717

T3 - Proceedings - 2023 IEEE International Conference on Big Data and Smart Computing, BigComp 2023

SP - 318

EP - 319

BT - Proceedings - 2023 IEEE International Conference on Big Data and Smart Computing, BigComp 2023

A2 - Byun, Hyeran

A2 - Ooi, Beng Chin

A2 - Tanaka, Katsumi

A2 - Lee, Sang-Won

A2 - Li, Zhixu

A2 - Nadamoto, Akiyo

A2 - Song, Giltae

A2 - Ha, Young-guk

A2 - Sumiya, Kazutoshi

A2 - Yuncheng, Wu

A2 - Kwon, Hyuk-Yoon

A2 - Yamamoto, Takehiro

PB - Institute of Electrical and Electronics Engineers Inc.

Y2 - 13 February 2023 through 16 February 2023

ER -

Park SJ, Kim J. DAG-GCN: Directed Acyclic Causal Graph Discovery from Real World Data using Graph Convolutional Networks. In Byun H, Ooi BC, Tanaka K, Lee SW, Li Z, Nadamoto A, Song G, Ha Y, Sumiya K, Yuncheng W, Kwon HY, Yamamoto T, editors, Proceedings - 2023 IEEE International Conference on Big Data and Smart Computing, BigComp 2023. Institute of Electrical and Electronics Engineers Inc. 2023. p. 318-319. (Proceedings - 2023 IEEE International Conference on Big Data and Smart Computing, BigComp 2023). doi: 10.1109/BigComp57234.2023.00065

DAG-GCN: Directed Acyclic Causal Graph Discovery from Real World Data using Graph Convolutional Networks

Abstract

Publication series

Conference

Keywords

Access to Document

Other files and links

Fingerprint

Cite this